I am working on a deep learning project to forecast Sudden Cardiac Death (SCD) using ECG data from PhysioNet. Specifically, I need to download and preprocess the following databases:

MIT-BIH Normal Sinus Rhythm (NSR) Database MIT-BIH Sudden Cardiac Death Holter (SCDH) Database

Goal:

• Download the full datasets programmatically in Python.
• Extract raw ECG signals and annotations from .dat, .hea, and .atr files.
• Preprocess the ECG data (denoising, R-peak detection, feature extraction).
• Train a deep learning model for SCD prediction

Attempted Approach: I tried using wfdb and physionet-datasets but ran into issues when downloading complete datasets. Here’s my current approach:

!pip install wfdb  

import wfdb    
record = wfdb.rdrecord('nsrdb/16265', pn_dir='nsrdb')  
annotation = wfdb.rdann('nsrdb/16265', 'atr', pn_dir='nsrdb')  

print(record.__dict__)  
print(annotation.__dict__)  

Issues & Questions:

1. How do I download ALL records from both databases in one go? (instead of manually specifying each file)
2. Are there better methods to bulk download PhysioNet datasets efficiently?
3. How can I extract full ECG signals along with annotations for preprocessing?
4. Any best practices for handling large ECG datasets in deep learning?